3D Camera Interaxial distance (Ideal distance is regarded as about the same as the human eyes)

Creative Scene Blocking for 3D Movies:

The aim of making technically good 3D movies, is so that there is no harm done to audiences. Harm can be done by any number of ways, top of the list being too much background or foreground parallax in a scene, thereby making it hard for the eyes to “fuse” parts of the 3D scene either in the distance or up close, and excessive “depth jump cuts” (when a preceding and followup scene are at different depths that force the eye to re-converge too quickly).

In this article we look at creative framing for medium to close-up action and a look at choices to be made in a 3D Camera Rig for shooting such scenes.

The Camera Rig of choice for closeup work, any Stereographer or DoP will tell you, is a Beam Splitter Rig, or Mirror box rig. This has been and still is the standard. The reason being a two way mirror lets any two cameras be aligned such that one camera shoots through an angled mirror, and the other located 90degrees either above the mirror or if the mirror is angled downward, then the camera is below (underslung 3D rig). This allows the two lenses of the cameras to have an interaxial (distance between the lens mid-points) of much less than human eye spacing if so needed for closeups.

When doing macro photography in 3D, this is highly desirable. When doing closeups for typical narrative movies, this is where there is room for argument and experimentation.

The video above can be seen in 3d on youtube, here. Aesthetics aside, (the video was shot with no live 3D monitoring or practical lighting), here is a breakdown of observations on the video:

Scene blocking in 3D for close-ups; The Positive parallax trap:

In this clip, the focus of attention is the hand moving the chess piece. There is nothing distracting in the background, just a bland floor. So… even though there is excessive positive parallax on the floor, would it hurt viewers eyes? The argument here is, that if there was something of interest to “fuse”, say some distinct objects or a pile of books…then audiences would automatically gravitate towards trying to “fuse” the book titles for example.

It would be even worse if depth of field was at play and the book titles were slightly out of focus. However, in this scene there is just a bland floor…Creative scene framing helps in closeup and even for general scenes when shooting a movie in 3D. This helps the Director achieve his vision for the story, by directing the audiences attention to the important part of the scene.

Another idea is to actually use Depth of field, (sometimes a bane for 3D) creatively, to completely blur out the background. This is for a later experiment and article. **update** here.

…It has to be noted that by it’s very nature, the rich visuals of a scene captured in Stereoscopic 3D, will make audiences detract from even the best Directors’ story telling ability. It’s human curiosity heightened by that suspension-of-disbelief that 3D movies create, without even trying. Creative scene blocking for 3D movies, helps achieve the best of both worlds.

More Thoughts and Observations:

• The video was shot with a pair of prosumer cameras, synced with LANC that have an interaxial of approx 3 inches and with no beam-splitter. The idea is to mimic new small cameras that are capable of high latitude such as a pair of Si2K minis.
• The distance to the scene being shot is approx 4 and a half feet. The images are later modified using techniques such as H.I.T (horizontal image translation) and can be refined further with re-converging software such as Nuke and Mystika.
• The argument for going with a non-beamsplitter rig is purely to experiment and see what kind of maneuverability is possible with a light weight rig in the hands of the DoP or Director to creatively block a scene and frame subjects – taking into account camera angles and weight of camera rig.
• One argument worth noting is… would a pair of modern cameras such as Iconix, Prosilica and Si2K minis, with physical profiles that allow side by side placement *AND* that offer dynamic range of approx 10 f-stops be comparable to Digital Cameras that are bulkier when slung on a beamsplitter rig but offer a few f-stops more?
• It has to be noted that you can lose up to 1 full f-stop or more with a beam splitter rig because of having to use ND filters and / or polarizers on one of the cameras to balance out the color shift and polarization issues of the mirror.

Distance to object is approx 4 1/2 feet. Shooting closeups in 3D with a parallel rig

3D Scene Framing and Invasion of Personal Space:

Something that may not be instantly clear from all the lessons learnt at Film School is that in a 3D movie, it’s not so easy to cheat Depth. In 3D movies the one thing that gets recorded is “Spatial information”. This is something that cannot be recorded in a 2D movie.

In scenes that feature OTS (over the shoulder shots), this brings up an interesting question… How close does the Subjective Camera have to be? Would we psychologically be “invading the personal space” of the main talent if it was a direct POV shot? or would we be actually bumped up against the back of the talent who’s shoulder is being framed?

In order to not break the stereo window in an OTS shot and using the rule of thumb of 1 – 30 for safe stereo (for every one inch of interaxial, nearest object is ideally 30 inches), if we managed a 3 inch interaxial, the nearest object would be 90 inches or 7.5 feet away.

However… what if we were to position the camera further behind, and zoom in just a bit (using lenses that do not increase the physical interaxial of the rig when placed side by side)…would we then be close enough to not break stereo windows and yet get good enough resolution on typical close-ups such as peoples faces, or other objects? This is worth experimenting with.

Above: Parallel non converged 3D rig scene framing after a zoom.

The video can be seen in 3D on Youtube, here. The camera rig is parallel and not toed-in or “converged”. It was shot “blind”, i.e with no on-set 3D monitoring to judge for depth, stereo window violations and more. With that in mind, and knowing that the rig has an interaxial of 3 inches and placed about 4  1/2 feet away, a zoom was applied. The video clip is not to show the act of zooming. It is to show that by setting a parallel rig at a distance, and zooming in for final framing, could the end result be as good as using a beam splitter rig?

Would there be much Spatial distortion or object size dis-orientation? (This can be seen on the carpet growing during the actual zoom) But the desired framing for the scene is when the zoom has ended.

With high resolution cameras and good ‘glass’, such scenes can be captured in 3D with lightweight 3D rigs These can then be steadi-cam mounted or hand-held giving the Director and DP endless creative choices. It is always important to monitor the live 3D signal for optimal results. This experiment was done “Hitchcock” style; back when video-assist was not available for 3D film making.

What is Spatial Resolution?

In 3D movie making the effect of Spatial Resolution on an audience’s mind and eyes has not yet been recorded. There is reason to believe that the sheer amount of spatial information presented in stereoscopic 3D to the eyes and brain, overcomes any pixel or f-stop resolution in a captured frame of video. Of course in an ideal world of 3d camera rigs, we would have both… and that would be one less excuse to want to shoot in 2D and convert to 3D.

For example: During a Medical operation that is filmed in 3D and shown either live to medical students or discussed later by a panel of Doctors, the difference is like night and day, when this footage is viewed in 3D versus a 2D version of the operation. The 3D version gives these students / doctors more spatial “depth” that aids their understanding of how to manipulate the delicate instruments and to comprehend the space between internal organs etc.

This “spatial resolution” will never get recorded in a 2D capture, even with a 5K or 10K sensor camera with 14 stops of dynamic range.

On a related note…DoPs and Directors need to give up the ghost on the belief that only the big bulky cameras have enough resolution or resolving latitude for feature film making. That was (probably) valid for 2D Cinema.  Even in 2D film making, the humble VDSLRs such as the Canon 5D MkII and 7D costing a couple of thousand dollars, now outshine and outperform the big giants. In upcoming firmware updates, it may be possible to record in full uncompressed resolution directly from these sensors over a simple HDMI cable and portable flash storage.

3D movie making is part Science and part Art:

Directors and DoPs thus need to come up with both technical AND creative ways to frame or block a scene, to reduce and/or, eliminate such discomfort. 3D movie making is both part science (technical) and part art. There is no one way of achieving the end result. The use of Stereoscopic Calculators, while helpful is not a guarantee that the end result will be aesthetically and artistically pleasing. Trying to create a 3D movie, solely by looking at data suggested by such Stereoscopic 3D calculators is akin to taking a “paint by numbers” approach to creating an art masterpiece.